Vous êtes sur la page 1sur 51





Prof. Dr. Andrei Avram

Octavia-Maria Șulea

JUNE 2012

1. Introduction. Traditional conjugational classes for Romanian.................................2

1.1. Latin inspired (infinitive-based) classification..................................................3
1.2. Additional subclasses for stem extensions [ez] and [esk].................................6
1.3. Disadvantages of an infinitive-based classification. Transition to finer-grained
2. Modern approaches................................................................................................10
2.1. Guțu-Romalo’s structural analysis..................................................................10
2.1.1. Stress shift in Romanian verbs..............................................................10
2.1.2. Affix inventory and ending patterns.....................................................13
2.1.3. Guțu-Romalo’s 10 conjugational classes..............................................17
2.1.4. Stem alternations...................................................................................19
2.1.5. Final paradigmatic classification..........................................................24
2.2. Problems raised by Guțu-Romalo’s analysis..................................................25
2.3. Feldstein’s segmentation of the Romanian verb.............................................30
2.4. Conclusions so far...........................................................................................30
3. Phonological processes in the verbal domain triggered by conjugation................31
3.1. Hiatus resolution.............................................................................................33
3.2. Regressive metaphony, diphthongization, and deletion..................................33
3.3. Palatalization...................................................................................................34
3.4. Feldstein’s derivation of indicative present forms..........................................36
4. Accounting for the alternations..............................................................................37
4.1. Reducing the ending sequences.......................................................................37
4.1.1. Verbs with theme vowel (stressed and unstressed) –e..........................37
4.1.2. Verbs with theme vowels –i and –î.......................................................39
4.1.3. Verbs with theme vowel –a .................................................................42
4.2. Proposing a final division................................................................................43
5. Conclusions and final remarks...............................................................................44


Traditionally, Romanian has received a Latin-inspired classification of verbs

into 4 (or sometimes 5) conjugational classes based on the ending of their infinitival
form. However, this infinitive-based classification proves itself inadequate in
describing and accounting for the irregularities in the paradigm, for the alternations
and syncretisms that occur. There have been, thus, numerous attempts throughout the
history of Romanian linguistics to give other conjugational classifications based on
the way the verb actually conjugates.
Lombard (1955), looking at a corpus of 667 verbs, combined the traditional 4
classes with the way in which the biggest two subgroups conjugate (one using the
suffix –ez, the other –esc) and arrived at 6 classes. Ciompec et al. (1985 in Costanzo
2011) proposed 10 conjugational classes, while Felix (1964) proposed 12, both of
them looking at the inflection of the verbs and number of allomorphs of the stem.
Guțu-Romalo (1968), analyzing a corpus of over 400 verbs, produced a list of 38
conjugational ending sequences, which she reduced to 10 conjugational classes.
For the purpose of machine translation, Moisil (1960) proposed 5 regrouped
classes of verbs, with numerous subgroups, and introduced the method of letters with
variable values, while Papastergiou et al. (2007) have developed a classification of
Romanian verbs in the indicative present from a (second) language acquisition point
of view, dividing the 1st and 4th traditional classes into 3 and respectively 5 subclasses,
each with a different conjugational pattern, and offering rules for alternations in the
Of the more extensive classifications, Barbu (2009) distinguished 41
conjugational classes for all tenses and 30 for the indicative present alone, covering a
whole corpus of over 7000 contemporary Romanian verbs, a corpus which was also
used by Dinu et al. (2012). Her classes were developed only on the basis of the
different conjugational endings, and the classification system did not take into account
the alternations occurring in the stem, while Dinu et al. (2012)’s experiment took into
account both stem alternation and ending sequences when modeling their 30
conjugational rules which they used to label their dataset. Finally, Feldstein (2004),
attempting to account for the syncretism between grammatical desinences of two
different paradigmatic slots displayed by Romanian, proposed a three-fold

segmentation of Romanian conjugational endings and identified a small number of
rules for the derivation of conjugated forms.
In what follows, I will look at both the traditional and modern analyses, and
attempt to give accounts for the occurring alternations and seeming richness of the
paradigm. Since Romanian verbs have different forms to express mood, tense, person,
number, gender, and voice, to simplify the analysis, I will limit the discussion to the
conjugation of Romanian verbs in the indicative present.

1.1 Latin inspired (infinitive-based) classification

As is the case for other Romance language, Romanian verbs have been
traditionally classified into 4 conjugational classes, which correspond to the ones in
Latin, as noted by Tiktin (1905: 98), and are based on the theme vowel surfacing as
the final vowel of the (present) infinitive form, Costanzo (2011: 96). Table 1 below
compares the Latin conjugational classes with Romanian.

Infinitive ending
Latin -ĀRE - ĒRE -ERE - ĪRE
Romanian -a -ea -e -i

Table 1. Conjugational classes in Latin and Romanian

Beside the verbs which end in one of these four theme vowels, Romanian also
has a number of verbs whose infinitive ends in –î. Since their conjugation is distinct
but argueably predictable from that of the verbs in the 4th class, Tiktin (1905: 98),
Lombard (1955), and others have traditionally taken these verbs to be part of the 4th
conjugation. Others, however, have argued for a separate analysis and postulated a 5th
class designed specifically for them. Costanzo (2011: 97), for instance, argues that the
split of –i and –î verbs into different classes can be justified, due to the predictability
of the conjugational pattern of –î verbs from –i verbs relying on Latin and phonology,
but unneccessary. The table below gives examples of verbs from each of the four
traditional Romanian conjugational classes.

Class Theme vowel Examples
I -a cita ‘quote”, invita ‘invite’, mânca ‘eat’, parca ‘park’
II -ea avea ‘have’, bea ‘drink’, vedea ‘see’, vrea ‘want’
III -e bate ‘beat’, crede ‘believe’, face ‘do’, merge ‘walk’
IV -i/ -î citi ‘read’, vorbi ‘speak’, hotărî ‘decide’, omorî ‘kill’

Table 2. The traditional conjugational classes in Romanian

The first conjugation is the richest and most productive, Avram (2001: 199),
followed in second place by the fourth conjugation (except the relatively small group
of verbs ending in –î, which is unproductive). When new verbs are created or
borrowed into the language, these two classes are the only ones which will accept
them, Constantinescu-Dobridor (2001: 143). The second and third conjugations are
limited to verbs inherited from Latin or Romance neologisms and are unproductive,
with the second class being the smallest of the four and appearing to die out,
Constantinescu-Dobridor (2001: 142). Specifically, a migration phenomenon in
Romanian has been noted by several linguists with respect to verbs from the second
conjugation which have acquired a third conjugation variant, as shown in (1) and (2):

(1) Diachronic transitions from II to III

a rămânea (<lat. remaneo) -> a rămâne ‘to remain’
a ținea (<lat. teneo) -> a ține ‘to hold’
a umplea (<lat. impleo) -> a umple ‘to fill’

(2) Transitions from II to III not yet accepted in Standard Romanian

a părea -> a pare ‘to seem, to appear’
a încăpea -> a încape ‘to fit’
a plăcea -> a place ‘to like’
a (pre)vedea -> a (pre)vede ‘to (for)see’
a (s)cădea -> a (s)cade ‘to (decrease/ subtract) fall’
a tăcea -> a tace ‘to shut up’

These transitions are natural and due to analogies and identical forms (i.e. the 1sg
form eu tac ‘I shut up’ of the verb a tăcea is identical to eu fac ‘I do’ of the verb a
face). Another already accepted migration, as observed by Avram (2001:199), is of
verbs which were initially from the third conjugation and have eventually come to be

accepted as first conjugation verbs (3). Some verbs from the third conjugation have
recently acquired a first conjugation variant which is however not yet accepted (4). A
final accidental transition, which is not accepted, can be noticed of fourth conjugation
verbs toward the third conjugation (5).

(3) Diachronic transition from III to I

a decerne -> a decerna ‘to award’
a precede -> a preceda ‘to precede’
a succeede -> a succeda ‘to succeed (smth.)’

(4) Accidental transition from III to I not yet accepted

a accede -> a acceda ‘to accede (to)’
a concede -> a conceda ‘to concede’

(5) Accidental transition from IV to III not accepted

a auzi -> a aude ‘to hear’
a absorbi -> a absoarbe ‘to absorb’
a împărți -> a împarte ‘to divide’

For the purpose of actually conjugating Romanian verbs, the four

conjugational classes act as guidelines which say what the standard ending pattern is
for a particular class. However, these classes say nothing about alternations which
may occur in the stem or in the ending during conjugation, as we will see further.
According to Papastergiou et al. (2007: 1) probably the most difficult Romanian tense
to learn as a foreign speaker is the indicative present. Table 3 shows the standard
ending patterns for each conjugational class, extracted from Avram (2001: 218).

Person I II III IV
Number –a –ea –e –i –î
1 sg – – – – –
2 sg –i –i –i –i –i
3 sg –ă –e –e –e –ă
1 pl –(ă)m –(é)m –(e)m –(i)m –(î)m
2 pl –(a)ți –(é)ți –(e)ți –(i)ți –(î)ți
3 pl –ă – – – –ă

Table 3. Standard ending patterns

A final thing to note is that verbs of the type a împerechea ‘to pair’, a veghea ‘to
watch over’, a iniția ‘to initiate’ are considered part of the first class, due to them
exhibiting the theme vowel a during conjugation: 2 pl. form of împerechea is
împerecheați and of iniția is inițiați.

1.2 Additional subclasses for stem extensions [ez] and [esk].

Apart from the standard ending patterns (and their variants which will be
discussed in another chapter), Romanian has two different classes of variable
functional infixes, [ez] and [esk], which attach to the stem of the verb before the
ending (desinence), as observed by Guțu-Romalo (1968: 151), Chițoran (2002a: 34),
Feldstein (ND: 9), and others. In the Romanian verbal domain, the infix [ez] attaches
to first conjugation verb stems, while [esk] attaches to fourth conjugation verb stems,
the latter surfacing in two variants: -esc- for verbs having the theme vowel –i and
-ăsc- for verbs having the theme vowel –î. The affix [ez] was inherited from the Latin
-IDI-, while [esk] from the Latin -ĪSC-/-ĒSC-, Costanzo (2011: 92). In Latin, when
the -ĪSC-/-ĒSC- affix was attached to a verb it had an inceptive function (verb +
[ESC] = to start verbing), however, this function was lost in Romanian. In the
nominal domain, on the other hand, the function of [ESC] added to a noun (that of
suggesting something prototypical of that noun) has been preserved in Romanian.
As stated previously, [ez] and [esk] are variable infixes, meaning they vary
with person and number, giving birth to different conjugational patterns for those first
and fourth conjugational verbs which receive these infixes. However, a separation
between the inflected (for person and number) form of the infix and the desinence can
be made, and, once it is, it becomes apparent that the desinence pattern which this
assumed separate group of verbs takes is the same as the one taken by verbs which do
not receive the infixes. Because of this, many linguists studying Romanian verbs have
extended the traditional ending pattern to the one collected in Table 4 from Avram
(2001: 218), where the classes of verbs taking either [ez] or [esk] are considered
subclasses of the 1st and 4th classes. Unlike this extended traditional analysis presented
in Table 4, Lombard (1974) and Lombard and Gâdei (1981), propose six
conjugations, where the two traditional subclasses of the traditional 1st and 4th
conjugations are considered separate classes, while the -i and -î subclasses are

Person I II III IV
& –a –ea –e –i –î
Number –ez +ez –esc +esc –ăsc +ăsc
1 sg – –ez – – – –esc – –ăsc
2 sg –i –ezi –i –i –i –ești –i –ăști
3 sg –ă –ează –e –e –e –ește –ă –ăște
1 pl –(ă)m –(ă)m –(é)m –(e)m –(i)m –(i)m –(î)m –(î)m
2 pl –(a)ți –(a)ți –(é)ți –(e)ți –(i)ți –(i)ți –(î)ți –(î)ți
3 pl –ă –ează – – – –esc –ă –ăsc

Table 4. Standard ending patterns with [ez] and [esk]

The class of verbs to which [ez] attaches (from here on +ez) is the largest
according to the quantitative results obtained by Dinu et al. (2012: 327), the second
largest being that of verbs receiving –esc (and not –ăsc). These two groups, +ez and
+esc, are usually the groups where newly created or borrowed verbs are included,
according to Avram (2001: 199), although which of the two infixes a neologism
receives is not a clear cut phenomenon, and competing variants coexist in the
contemporary language, observes Costanzo (2011: 116).

1.3 Disadvantages of an infinitive-based classification. Transition to finer-

grained classifications

If we take conjugational class to mean a conjugational pattern which is

followed by all verbs pertaining to that class in the exact same way, as Costanzo
(2011: 19) does, and if we then turn to analyze Romanian verb paradigms, we
inevitably end up seeing that only four (in the traditional classification) or eight (in
the extended traditional) classes are far from describing this verbal inflectional system
in its entirety.
On the one hand, the ending patterns, listed in Table 4 of the previous
subsection and representing the extended traditional conjugational classification, are
not all the patterns which occur in the Romanian verbal domain. For instance, shown
in Table 5, the verbs a copia ‘to copy’, a da ‘to give’, a bea ‘to drink’ a scrie ‘to
write’, and a oferi ‘to offer’, although belonging to one of the four traditional classes,

based on their infinitive, do not conjugate exactly according to any of the patterns
displayed by these classes in Table 4.

a copia a da a bea a scrie a oferi

copiez dau beau scriu ofer
copiezi dai bei scrii oferi
copiază dă bea scrie oferă
copiăm dăm bem scriem oferim
copiați dați beți scrieți oferiți
copiază dau beau scriu oferă

Table 5. More conjugational patterns

On the other hand, Romanian has quite a rich inflectional morphology, which
means that it certainly does not contain only regular verbs. There are plenty of verbs
which display alternations in the stem (stem allomorphy), meaning that their stem
surfaces differently throughout the paradigm. Guțu-Romalo (1968: 272) divides verbs
according to stem alternation into 4 classes: verbs with invariable stem (regular),
verbs with partially variable stem (partially irregular), verbs with partially aberrant
variation in the stem, and verbs which display suppletive allomorphy (irregular), the
largest of these four classes being the first two. A good thing about verbs which have
partially variable stem is that their alternation can be predicted phonologically
something which will be discussed in the following chapters. Table 6 below displays
an example for each of these classes of verbs.

Type of stem Infinitive form Indicative present conjugation

invariable a dansa dans-ez, dans-ezi, dans-ează,
‘to dance’ dans-ăm, dans-ați, dans-ează
partially variable a tresălta tresalt, tresalți, tresaltă
‘to start/exult’ tresăltăm, tresăltați, tresaltă
partially aberrant a mânca mănânc, mănânc-i, mănânc-ă
‘to eat’ mânc-ăm, mânc-ați, mănânc-ă
suppletive a avea am, ai, are,
‘to have’ avem, aveți, au

Table 6. Classification according to stem variation

Another reason to consider the traditional classification as being under-
informative is pointed out by Feldstein (2004), and has to do with syncretism patterns
within paradigms, namely, the fact that verbs from the same conjugation display
different syncretism patterns. This is exemplified in Table 7.

Infinitive 1sg 2sg 3sg 1pl 2pl 3pl

invita ‘invite’ invit inviți invită invităm invitați invită
tăia ‘cut’ tai tai taie tăiem tăiați taie
da ‘give’ dau dai dă dăm dați dau

Table 7. Different syncretism patterns in the same conjugation

These facts show that a classification of Romanian verbs only according to

theme vowel (which surfaces in the infinitive “dictionary” form as the final vowel) is
insufficiently descriptive of their paradigms. The contemporary data, collected by
Barbu (2009) and analyzed from a computational point of view by Dinu et al. (2012),
strengthens previous linguistic analysis according to which a simple inventory of
paradigms is unproductive and near impossible to make, and shows that new or more
criteria need to be added in the classification of Romanian verbs. In what follows,
finer grained classifications along with their criteria will be presented and analyzed
from a theoretical point of view.


2.1 Guțu-Romalo’s structural analysis

As touched upon in the first chapter, Romanian has a rich inflectional

morphology. This richness burdens morphology acquisition for a foreign learner but
also burdens a linguistic attempt to describe the system and find the right criteria to
classify it. The structural analysis of Guțu-Romalo (1968: 275) has shown that the
verbal inflectional system is arguably more complex than the nominal one. Her study
looked at over 400 verb paradigms and classified them based on their stress pattern,
endings pattern, and stem allomorphy.
After she analyzed the affixes attached to the stem and created a large
inventory of ending patterns, she attempted to extract a conjugational classification,
reducing the 38 identified sequences to 10 conjugational classes, based on several
types of homonymy between the affix sequences. She then analyzed stem alternations
and attempted to unify her conclusions about stress, conjugational class, and stem
alternation in the Romanian verbal domain under one “paradigmatic classification”.
Let us go through her structural analysis and see how this was obtained.

2.1.1 Stress shift in Romanian verbs

She first argues that the study of stress placement and movement in the
Romanian verbal domain should precede any analysis regarding stem alternations and
ending patterns. According to her, stress may fall either on the stem or on the ending,
having enough mobility to shift from one part to the other within the same paradigm,
Guțu-Romalo (1968: 152). When stem and ending are taken globally, the author
identifies three stress patterns for the indicative present:
A. stress falls on the stem for the 1st, 2nd, and 3rd person singular and 3rd
person plural, and shifts on the ending for 1st and 2nd person plural;
B. stress is maintained on the stem for all forms
C. stress is maintained on the ending for all forms
She claims that A applies to verbs whose infinitives end in –a, –ea, –i, and –î, B
applies to verbs ending in –e, and C applies to verbs ending in –a (+ez), –i (+esc), and
–î (+ăsc). However, monosyllabic verbs like a da, which don’t have where to shift the
stress since they don’t acquire another syllable when conjugated in the indicative
present, seem to fit, stress pattern wise, either under C or B (depending on what
analysis of constitutive parts they receive), although they don’t respect the conditions
proposed by Guțu-Romalo for this classification. In (6), a da ‘to give’ and a lua ‘to
take’ are conjugated in the indicative present with phonological representations.

(6) a da a lua
[daw] [jaw]
[daj] [jej]
[dǝ] [ja]
[dǝm] [lwǝm]
[dats ] [lwatsj]
[daw] [jaw]

Example (6) shows that an analysis of a verb’s constitutive parts also needs to come
into play when describing stress placement and stress mobility and that there are
definitely examples of verbs which do not follow Guțu-Romalo’s general
classification of verbal stress pattern.
When stem and ending are not monosyllabic, and when the stress falls on the
stem, Guțu-Romalo (1968: 156) claims that, for trisyllabic stems, the stress can either
fall on the final or penultimate syllable, but never on the first syllable. For disyllabic
stems, the stress can fall on either of the two syllables. According to her study, the
most numerous are verbs with bisyllabic final stressed stem. Chițoran (2002a: 56)
further notes that, out of the 803 basic verbs counted by Juilland et al. (1965),
penultimate stress on the stem (root) occurs only in 81 verbs, with 62 belonging to the
first conjugation and 16 to the fourth, which leads her to analyze these verbs as
lexically marked exceptions.
When the stress falls on the ending, presupposing that it is phonetically
realized (meaning a non-zero morpheme), it always falls on the first syllable, if it is
bisyllabic (it can at most be bisyllabic). Chițoran (2002a: 57-58), on the other hand,
takes the theme vowel and the affixes –ez and –esc to be part of the stem, thus finding
that the prosodic word has nothing to do with verb inflection, due to stress never
falling on the inflectional ending, but falling either on the theme vowel, –ez or –esc,
when they are expressed, or on the final/ penultimate syllable of the stem, when they
are not expressed. This interpretation contradicts Guțu-Romalo (1968: 157)’s
statement according to which, when the stem is stressed, the stress will always stay on
the same syllable, exemplified in (7), where the stem is between square brackets.
(7) a. Guțu-Romalo (1968: 144)’s segmentation:
[sp ri]-i sau [sp r]- i
[sp ri]-i sau [sp r]- i
[sp r]- ie
[sper]- i m
[sper]- i tsi
[sp r]- ie
b. Chițoran (2002a: 55)’s segmentation:
[adún]- j
[adun- ]-ts j

Due to differences in segmentation, the stem in (7a) looses main stress in favor
of the ending for the 1st and 2nd person plural, but when the stem is stressed it’s
always stressed on the same syllable. In (7b), however, stress remains on the final
syllable of the stem, shifting together with the final syllable, which changes for the 1st
and 2nd person plural. As was the case with example (6), we once again see that the
analysis of stress placement and stress mobility in the Romanian verb is dependent on
the analysis of its constitutive parts. Another thing to note is the different
phonological interpretation of certain sounds, Guțu-Romalo not offering an account
for her choice of representation for the final i in sperii ‘I scare’.
Finally, Guțu-Romalo (1968: 156) identifies 4 classes of verbs based on which
(of the three) types of stress mobility they have for each tense. For the indicative
present, these 4 regroup into 3 classes quoted in the table below.

Class Stress pattern Verbs ending in

I A -a, -î, -i, -e, -ea
II C -a (-ez), -i (-esc), -î, (-ăsc)
III B -e

Table 8. Classification of verbs according to type of stress mobility

2.1.2 Affix inventory and ending patterns

As mentioned in the previous subsection, Guțu-Romalo (1968: 167) considers

verbal inflectional endings to be composed of a variable suffix, which can be
phonetically realized in up to 4 forms and is almost always realized for the 1st and 2nd
person plural, and a desinence (which displays person and number) and attaches to
this suffix, when it surfaces, or directly to the stem.
For the indicative present, she identifies a total of 15 types of suffixes, which
are given in Table 9. For the desinences, Guțu-Romalo (1968: 169) takes each person
and number and lists their phonetic realization, as gathered in Table 10. She attempts
to combine the two lists in order to classify verbs based on inflectional patterns and
identifies 38 ending sequences, Guțu-Romalo (1968: 195-197), which are reduced to
10 conjugational classes, Guțu-Romalo (1968: 203-205).

No. Realization Examples

1 [- , - ,] ara ‘plow’, da ‘give’, săpa ‘dig’, sta ‘stay’
2 [-i , -i , ] apropia ‘bring near’, mângâia ‘caress’, speria ‘scare’
3 [- , - , ] continua ‘continue’, lua ‘take’
4 [- , - , - z, -e z] lucra ‘work’, dansa ‘dance’
5 [- , - , - z, - z] veghea ‘guard’:
6 [-i , -i , -i z, -i z] sublinia ‘emphasize’
7 [- , - , - z, -e z] crea ‘create’
8 [-î, ] coborî ‘descend’, omorî ‘kill’, doborî ‘knock off’
9 [-î, - sc, - șt] hotărî ‘decide’, rî ‘hate’
10 [- , ] fugi ‘run’, veni ‘come’, sui ‘climb’
11 [- , - sc, - șt] goni ‘run’, răpi ‘kidnap’, luci ‘glitter’
12 [- , -i sc, -i șt] alcăt i ‘form’, construi ‘construct’, dăr i ‘gift’
13 [- , ] vedea ‘see’, părea ‘seem’
14 [-e, ] arde ‘burn’, prinde ‘loose’, rupe ‘tear’, scrie ‘write’
15 [] ști ‘know’

Table 9. Guțu-Romalo’s verbal suffix series

As will be discussed in a following section, the disadvantage of an inventory

as the one in Table 7 is that it seems to be based more on orthographical form, rather

than phonological form. As the author herself notes later on, Guțu-Romalo
(1968: 200), some of the “suffix patterns” can be shown to derive, phonologically,
from others, yet she never goes into specifying how. For instance, pattern 5 looks very
much like 6, and, when the paradigm of the representing verb a veghea ‘to watch
(over)’ is analyzed, it seems it would rather fit into pattern 6 than 5. The comparison
between how the verb is written in Romanian and how the author segments and
interprets its paradigm is shown in (8).

(8) a veghea, orthographically: vs. a veg’a, where [g’] = ghe:

veghez, veghez-i, vegheaz-ă veg’-ez, veg’-ez-i, veg’-az-ă
veghe-m, veghea-ți, vegheaz-ă veg’-e-m, veg’-a-ți, veg’-az-ă

The author does mention the phonological context (preceding sound or

syllable) in which each of these 15 patterns or sequences tend to appear, however, she
never attempts to account phonologically for the selection of one pattern over another
by a particular type of verb, although she does mention that in some cases the
selection is phonological. Chițoran (2002a)’s segmentation (discussed in 2.1.1) seems
to induce a smaller number of these so called suffix patterns. This will, however, be
analyzed later.

Number Person Forms

1st d11=[-], d12=[-i], d13=[-u], d14=[-w], d15=[-m]
Singular 2nd d21=[-i ], d22=[-i], d23=[- ]
3rd d31=[-ă], d32=[-uă], d33=[-e], d34=[-ie], d35=[-]
1st d4=[-m]
Plural 2nd d5=[- tsi ]
3rd homonymous with 1st or 3rd person singular forms

Table 10. Verbal desinences identified by Guțu-Romalo

Table 10 shows the various desinences gathered by the author from the corpus.
For the second person singular and plural desinences, d21 and d5, i is taken to be a
non-syllabic realization of the phoneme i, which is known to cause palatalization in
the preceding consonant, Chițoran (2002a). In this second listing, Guțu-Romalo goes
into more detail to describe the context in which these endings occur. The context for
each of the alternating desinence is as follows:

 d11 for verbs with stems either ending in a consonant or consonant cluster (with
the exception of stop+liquid) and no suffix (i.e. fug-- ‘I run’), or with suffix
(i.e. lucr-ez- ‘I work’);
 d12 for verbs with stems ending in a vowel and no suffix: ta--i ‘I cut’;
 d13 stem ending in a stop+liquid or liquid+liquid cluster: umpl--u ‘I fill’;
 d14 stem ending in u and no suffix (i.e. continu--u ‘I continue’) or monosyllabic
aberrant stem ending in a vowel (i.e. scri--u ‘I write’);
 d15 only for the verb a avea ‘to have’: am ‘I have’;
 d21 all stems ending in a consonant (cluster), except stop+liquid and liquid+liquid,
with suffix (i.e. lucr-ez-i ‘you work’) or without suffix (i.e. arz--i ‘you burn’);
 d22 stem ends in a stop+liquid or liquid+liquid cluster and receives no suffix:
url--i ‘you yell’;
 d23 stem ending with a vowel and having no suffix: scri--i ‘you write’
 d31 all verbs from the first conjugation whose stems end in a consonant (urc--ă
‘s/he climbs’), all verbs with the infinitive in –î and no suffix (coboar--ă ‘s/he
descends’), and some verbs with the infinitive in –i and no suffix (ofer--ă ‘s/he
 d32 first conjugation verbs with stems ending in –u: continu-- ă ‘s/he continues’;
 d33 verbs in –i, except those in d31 (i.e. fug--e ‘s/he runs’, ghic-eșt-e ‘s/he
guesses’), verbs in –î with the suffix –ășt (i.e. hotăr-ășt-e ‘s/he decides’), and
verbs from the 2nd and 3rd conjugation (i.e. );
 d34 verbs whose stems end in a vowel (except those in d32): [skri--ie] ‘s/he
 d35 restricted to a few aberrant verbs: ia- ‘s/he takes’, bea- ‘s/he drinks’, vrea-
 ‘s/he wants’.
After the inventories presented in Table 9 and 10 are made, Guțu-Romalo
(1968) combines the two classifications and identifies 38 conjugational patterns or
sequences, which extend the (5 or 8) traditional ending patterns discussed in the first
chapter. Since the list is quite long, it is quoted in appendix A.
Following the lengthy inventory of conjugational patterns, the author discusses
several types of homonymy between suffixes of different tenses and between
desinences within paradigms. Through these observed types of homonymy, she is able

to narrow down the 38 patterns to 10 conjugational classes, which will be presented in
the following subsection.
Since our current analysis is restricted to the Romanian indicative present, the
homonymy between desinences occurring within the boundaries of this tense is
interesting to observe. She identifies three types of desinence homonymy for the
indicative present:
1. between 3rd person singular and plural;
2. between 1st person singular and 3rd person plural;
3. between 1st person singular and 2nd person singular.
Out of the 38 conjugational classes, Guțu-Romalo (1968: 198) found that
almost each class displayed one of these 3 types of desinence homonymy. There was
one class (namely 14) that displayed both type 1 and type 3 homonymy, while there
were two other classes (34 and 35) that did not display any of these three types. In (9)
below, the three types of homonymy are exemplified.

(9) a. type 1 & 3: a sui ‘to climb’

s iți
b. type 2: a rupe ‘to tear’
r peți

The underlying reason for this phenomenon is not discussed, the author
limiting herself to simply recording it. The discussion is later picked up by several
other linguists, with one of the most recent analysis belonging to Feldstein
(2004: 179), who argues that these three paradigm phenomena in the indicative
present are a form of phonologically conditioned syncretism.

2.1.3 Guțu-Romalo’s 10 conjugational classes

In her process of reducing the 38 ending sequences to 10 inflectional

(conjugational) classes, the author of this structural analysis identifies three causes
that determine complexity (expressed through a great number of different ending
sequences) in the inflectional domain of the Romanian verb:
1. phonetically conditioned allomorphy;
2. morphologically conditioned allomorphy;
3. lexical allomorphy.
With this observation at hand, Guțu-Romalo (1968: 200-203) unites ending patterns
(from her inventory of 38) when their differences are solely based on phonological
reasons, and identifies distinct conjugational classes, when their differences cannot be
accounted phonologically, but morphologically. Thus, the main criterion by which she
identifies a distinct conjugational class is morphological. Table 11 below describes
each conjugational class for the indicative present, where di represents the desinence
for the i-th person (i.e. d4 refers to 1st person plural, d6 to 3rd person plural, etc.) and
the quoted affixes are the indicative ones which are shared with other tenses. The
series column shows which ending sequences are represented by each class.

Cls. Series Affix Desinence Examples

I 1-6 {- -} d3 = d6 = {-ă} ara ‘plow’, apropia ‘draw near’
II 7-11 {- ~ z-} d3 = d6 = {-ă} lucra ‘work’, sublinia ‘emphasize’
III 12 {-î- } d3 = d6 = {-ă} coborî ‘descend’
IV 13, 14 {- -} d3 = d6 = {-ă} acoperi ‘cover’, sui ‘climb’
V 15, 16 {- -} d1 = d6 = {-} sări ‘jump’, fugi ‘run’
VI 17-20 {- ~ sc} d1 = d6 = {-} isprăvi ‘finish’, hotărî ‘decide’
VII 21, 22 {- -} d1 = d6 = {-} părea ‘seem’, tăcea ‘shut up’
VIII 23-25 {-e-} d1 = d6 = {-, -u} umple ‘fill’, face ‘do’
IX 26-28 {-e-} d1 = d6 = {-, - } prinde ‘catch’, scrie ‘write’
X 29, 30 {-e-} d1 = d6 = {-} rupe ’tear’, coace ‘bake’

Table 11. Guțu-Romalo’s conjugational classes

The first thing we notice, looking at Table 11, is that only the ending
sequences 1-30 are represented by these 10 classes. The author explains that she takes
these first 30 sequences to represent regular verbs, Guțu-Romalo (1968: 203).
Another thing to notice is that, at least according to the information captured in
the table above, classes VIII to X seem not to differ in indicative present affix and
desinence syncretism. Actually, when looking in Appendix A at series 23-30, the only
differences in the indicative present that these ending patterns display is in 1st person
singular and 3rd person plural desinence. Namely, sequences 24 and 28 are almost the
same in the indicative present but differ from all the others (which are exactly the
same among each other) in that d1 = d6 = [u] for sequence 24 and d1 = d6 = [ ] for
sequence 28. This difference is taken by the author to be phonological, and that is why
24 is grouped with 23 and 25 (which display d1 = d6 = ) in class VIII and 28 is
grouped with 26 and 27 (which also display d1 = d6 = ) in class IX.
The reason why patterns 23-30 are not grouped together into one single
conjugational class appears to be a more poignant difference in their other tense
forms. Namely, the three class difference identified by the author (meaning 23-25 vs.
26-28 vs. 29, 30) becomes apparent only when looking at the participle forms, which
group together exactly according to this classification (23-25: [- + t]; 26-28: [ + s];
29, 30: [ + t]). However, if we take simple perfect or pluperfect (which is always
deductable from the simple perfect) forms to be the grouping criterion for these
sequences, we end up grouping them differently: sequences 23 to 25 group together
under the pluperfect form [- + se-], while sequences 26 up to 30 group under the
form [-s + se-], which implies only two different classes. If we take the subjunctive
present as a grouping criterion, then we arrive at 3 classes but differently divided than
the ones identified by Guțu-Romalo (1968: 202). Given these facts, I see no reason to
assume one tense as a division criterion over the other.
Since our analysis is restricted to the indicative present, assuming all other
reasons presented by the author in her 10-fold classification are valid, the differences
between ending patterns in other tenses should not be relevant to a classification of
ending patterns in the indicative present. Thus, the ending sequences represented by
classes VIII to X should be taken to represent one conjugational class, when looking
only at the indicative present.

2.1.4 Stem alternations

A classification of Romanian verbs according to stem alternation slightly

simpler than the one presented in section 1.4 is a three-fold one, as identified by
Guțu-Romalo (1968: 211), where verbs can either be:
1. verbs with a constant stem throughout their entire paradigm (i.e. rupe
‘tear’, ara ‘plow’); they are called regular;
2. verbs whose stem is partially variable (i.e. cere ‘ask’); these are called
partially irregular;
3. verbs whose stem has suppletive allomorphy or fully irregular verbs (i.e. a
fi ‘to be’).
Due to the second class being the most problematic, Guțu-Romalo (1968: 211-248)
begins by first looking at these verbs, and attempts to account for the consonantal and
vocalic alternations that their stems go through. Table 12 below collects the consonant
alternations identified by Guțu-Romalo to occur in Romanian verb stems.

No Alternation Example
1 [k] / [ʧ] a călca ‘to step’, calci [kalʧ i] ‘(you) step’
2 [g] / [ʤ] adaug ‘(I) add’, adaugi [adauʤi] ‘(you) add’
3 [t] / [ʦ] aștept [aʃtept] ‘(I) wait’, aștepți [aʃtepʦi] ‘(you) wait’
4 [d] / [z] văd [vǝd] ‘(I/they) see’, vezi [vezi] ‘(you) see’
5 [s] / [ʃ] cos ‘(I/they) sew’, coși [koʃ i] ‘(you) sew’
6 [s (t)] / [ʃ (t)] gust ‘I taste’, g ști [guʃti] ‘(you) taste’
7 [(ʃ) k] / [(ʃ) t] m șc ‘I bite’, m ști ‘you bite’
8 [(s) k] / [(ʃ) t] nasc ‘(I/they) give birth’, naști [naʃti] ‘(you) give birth’
9 n/ rămân ‘(I/they) remain’, rămâi [rǝmɨj] ‘(you) remain’
b/ fierb ‘(I/they) boil’, fier-se ‘s/he boiled’
10 [k] / [ʧ] /  d când ‘leading’, duci [duʧ i] ‘you lead’, du-se ‘s/he lead’
[g] / [ʤ] /  spărgând ‘breaking’, spargi ‘you break’, spar-se ‘he broke’
[t] / [ʦ] /  a scoate ‘to remove’, scoți ‘you remove, scoa-se ‘he
[d] / [z] /  removed’
11 [k] / [ʧ] / [p] coc-ând ’baking’, coci [koʧ i] ’(you) bake’, cop-sei ‘I baked’
12 [g] / [ʤ] / [p] înfig-ând ‘sticking’, înfige [ɨnfiʤe], înfip-se ‘stuck’

Table 12. Consonantal alternations identified by Guțu-Romalo

The final part of the alternations under 11 and 12 (involving [p]) only occurs
in other tense forms, so these alternations are excluded from our analysis as they are
irrelevant to the indicative present and, since the first part of these alternations is
already captured in 1 and 2, alternations 11 and 12 will be completely ignored in our
analysis. A similar thing happens to the final component in the alternations under 10.
Namely, the final consonant in the stem (/k/, /g/, /t/, /d/, and /z/) is deleted only in
other tenses (simple perfect, pluperfect, and participle), while the indicative present
triggers the corresponding alternations already described in 1-4. Finally, the second
alternation in 9 (b/) is isolated to the verb a fierbe, which does not display the
deletion in the indicative present. Alternations 6 to 8 all seem to be related somehow
to alternation 5 and Chițoran (2001a: 192-195) is able to account for all these in a
similar way.
Regarding the orientation of these alternations (i.e. which form comes first/ is
underlying), Guțu-Romalo (1968: 214) takes it to be arbitrary and without real
consequence to the description. She chooses the element in the 1st person singular
form to represent the first element in an alternation by arguing that in most cases this
form has null ending so would most likely display the uninflected (underlying) form
of the stem. Chițoran (2002a: 43) assumes that the first person singular present
indicative form does have an inflectional marker, an underlying /-u/ which is deleted
when preceded by only one consonant. However, even when it surfaces, the
realization of this person marker does not affect the final element of the stem so it
does not trigger alternation, which means that the assumption that the 1st person
singular indicative present form displays the underlying form of the stem is a safe one.
According to Guțu-Romalo (1968: 214), consonant stem alternations in the
Romanian verb paradigm always and without exception occur at the juncture between
stem and the various affixes which get attached during conjugation, are restricted to
one per stem, and are not influenced by stress. The first most often cause for the
occurrence of the second element of the alternations listed above is the non-syllabic
surface form of the 2nd person singular marker –i, which triggers palatalization of the
preceding consonant, Chițoran (2002a: 173), and triggers either one of the alternations
listed above (except b/). In this case, the occurrence is phonologically conditioned.
Some of the alternations (2-5) can also appear when the stem is followed by a
vocalic realization of –i, in which case the alternation can be argued to not be just
phonological, but also morphological, since vocalic –i characterizes verbs with the

infinitive in –i. Alternation 3 is argued by Guțu-Romalo to be lexically conditioned
for a small number of verbs whose stem ends in -s and infinitive in –i, which receive
the same ending pattern as other verbs displaying this stem alternation, but have an [s]
instead of a [ʃ] preceding the theme vowel in their infinitive, as shown in (10).

(10) [s], when followed by [ ]:

a. becomes [ʃ]: i s ‘(I/they) exit’, ieș ‘to exit’ (alternation 3, [s]->[ ʃ]);
b. remains [s]: mir s ‘(I/they) smell’, miros ‘to smell’ (no alternation).

The two types of verbs represented in (10) display alternation 3 in different

environments: ieși displays 3 when followed both by syllabic and nonsyllabic –i,
whereas mirosi only in the context of nonsyllabic manifestation of the 2nd person
singular marker –i. For alternations 1, 2, 7, and 8, the second element of the
alternation can also appear after the 3rd person singular marker –e which is always
syllabic and phonologically conditions the alternation, Guțu-Romalo (1968: 216-217).
The second type of stem alternation involves the stem’s vocalic elements,
which, according to Guțu-Romalo (1968: 223) and as discussed in 2.1.1, can be
stressed or unstressed. This mobility of stress leads to the following division of
vocalic alternations:
1. vocalic alternations which maintain stress (Table 13)
a. within stressed syllable
b. within unstressed syllable
2. vocalic alternations which involve stress loss (Table 14).

Syllable Alternation Examples

ă/a ag ț ‘(I) hang’/ ag ță ‘(s/he/they) hang’
e/a ș d ‘(I/they) sit’/ ș de ‘(s/he) sits’
î/i v nd ‘(I/they) sell’/ v nde ‘(s/he) sells’
stressed e/ea al rg ‘(I) run’/ ale rgă ‘(s/he/they) run’
o/oa r d ‘(I/they) chew’ / roade [rw de] ‘(s/he) chews’
ă/e/a sp l ‘(I) wash’/ sp li ‘(you) wash’/ sp lă ‘s/he/they
unstressed ă/e păr ‘(I) defend’/ peri ‘(you) defend’

Table 13. Vocalic alternations which maintain stress

Alternation Examples
/ă b g ‘I insert’, b gă ‘(s/he/they) insert’ / băg m ‘(we) insert’
/e v n ‘(I/they) come’ / ven m ‘(we) come’
/ us c ‘(I) dry’, us ci ‘(you) dry’, usc m ‘(we) dry’
/o /u p rt ‘(I) wear’, poartă [pw rtǝ] ‘(s/he/they) wear’, purt m ‘(we)

Table 14. Vocalic alternations with stress loss

The phonological conditioning is more proeminent here than it is in

consonantic alternations, while the morphological conditioning is indirect. Appendix
B quotes the interactions identified by Guțu-Romalo (1968: 234) between consonantal
and vocalic alternations. Unlike the consonantal ones, vocalic alternations rarely
occur on the final position of the stem, due to stems typically ending in a closed
syllable (i.e. having a coda) in Romanian, which means that the final stem position is
usually occupied by a consonant.
Since the interaction between the two main types of alternations is quite
complex and a stem can undergo several alternations during conjugation,
Guțu-Romalo (1968:235) takes the various results of alternation in a stem to represent
allomorphs of that stem which should be listed separately from a conjugational
classification. She even goes on to classify verbs according to how many allomorphs
their stem has, identifying verbs with at most 6 stem allomorphs within their whole
paradigm (11) and at most 4 allomorphs within the indicative present (12).

(11) tr g ‘I pull’
trăg(înd) ‘pulling’
tr g(e) [tr ʤ-e] ‘(s/he) pulls’
trăg(eam) [trǝʤ-e m] ‘(I) was pulling’
tra(se) ‘(s/he) pulled’
tră(sei) ‘(I) have pulled’

(12) port ‘(I) wear’

porț-i ‘(you) wear’
poart-ă ‘(s/he/they) wear’
purt-ăm ‘(we) wear’

These alternations usually happen at the joncture between stem and
conjugational affixes and are caused by affixation, which means that there is a part of
the stem which doesn’t suffer alternation and is usually invariable throughout the
paradigm. Because of this, Guțu-Romalo (1968: 235) takes verbs who suffer a
modification of the first element of the stem to have an aberrant variation and
classifies them outside of the partially irregular verb group.
Thus, a partially irregular verb, according to Guțu-Romalo (1968), means a
verb which has up to 4 stem allomorphs in one tense or up to 6 forms in the paradigm
and an invariable part at the begining of the stem. When the stem has more than 6
diffrent forms throught the paradigm and these forms do not share an invariable part,
then the stem is said to go through supletion and the verb is deemed fully irregular.
Guțu-Romalo (1968: 249-252) identifies 2 fully irregular verbs, a fi and a lua (13),
and 6 partially irregular verbs presenting aberrant paradigms: mânca ‘eat’, da ‘give’,
sta ‘stay’, bea ‘drink’, vrea ‘want’, and avea ‘have’. She notices certain similarities
between the verbs da, sta, bea, and vrea, but claims the variations in each paradigm
do not allow a reunion of these verbs under one rule, Guțu-Romalo (1968: 252).
However, the unaccountable variations she identifies do not occur in the indicative
present, in which tense these verbs seem to follow two patterns, as shown in (14),
with the exception of vrea which exhibits aberrant stem in the 3rd person plural.

(13) a fi ‘to be’ a lua ‘to take’

sunt iau
ești iei
este ia
suntem l ăm
s nteți l ați
sunt iau

(14) a da a sta vs. a bea a vrea

dau stau beau vreau
dai stai bei vrei
dă stă bea vrea
dăm stăm bem vrem
dați stați beți vreți
dau stau beau vor

Guțu-Romalo (1968: 240-247)’s classification based on how many stem
allmorphs a Romanian verb has also includes infromation about which person and
number forms share the stem (i.e. stem syncretism information) which leads to further
subdivisions in this classification. The reduced (to the indicative present) and thus
reordered version of her classification is given Appendix C.

2.1.5 Final paradigmatic classification

After classifying Romanian verbs according to stress mobility (Table 8),

ending patterns (Appendix A, Table 11), and stem alternations (Tables 12, 13, and 14,
Appendix B), Guțu-Romalo (1968: 261-264) attempts to reunite this information
under one paradigmatic classification. She first very straightforwardly intersescts the
stress and general ending pattern classifications, obtaining the distribution of her 10
conjugational classes according to stress mobility, which is quoted in Table 15.

Stress mobility class Conjugation


Table 15. Distribution of conjugations according stress classes

After this, she proceeds to include stem alternation within the conjugational
classification by distributing the stem allomorphy subtypes, which she identified in
Guțu-Romalo (1968: 240-247) and which I reduced and regrouped in Appendix C,
according to her 10 conjugational classes. The resulting classification (even when it is
done with the reduced and regrouped versions for the indicative present) is
cumbersome and much too close to a simple enumeration, as the author herself notes
in Guțu-Romalo (1968: 265).
However, she insists that information regarding stem allomorphy must not be
overlooked when attempting to create a description of Romanian verbs in such a way
as to be able to fully reconstruct their paradigms from the given information. This
leads her to look for a diffrent approach to characterizing the stem, and settles on one
inspired by Moisil (1960)’s variable letter solution, with her version reliying on
phonetic representations instead of written form.

The solution proposed by Moisil (1960) and revisited by Guțu-Romalo (1968)
involves using a symbol formed out of a letter and a subscripted digit in the notation
of a stem to suggest what type of alternation that letter/sound will get and in what
context. The example below extracts Guțu-Romalo’s use, with si meaning the i-th
surface stem in the indicative present (i.e. s4 means 1st person plural stem):

(15) a cân a ‘to sing’, where is:

 ț in and t in all other forms of verbs from the 1st, 2nd, and 4th conjugation;
 ț in , , and and t in all the other forms of verbs from the 5th conjugation.

This notation requires the prior definition of these variable letters (each indice
representing one type of alternation). Guțu-Romalo (1968:266-271) thus lists 30 of
these variable letters with their context for all tenses.

2.2 Problems raised by Guțu-Romalo’s analysis

As noted by Avram (1969b: 557) and discussed in 2.1.1, a first problem in

Guțu-Romalo’s analysis arises from her unexpected and ill-founded choice of verb
segmentation. Besides taking the affixes that tie stem and desinence together to be
part of the ending rather than part of the stem, a choice which leads to more ending
patterns as briefly discussed in section 2.1.1, Guțu-Romalo also wrongly identifies the
stem in some cases. Example (16) below quotes the two possible “stem-ending”
segmenations which Guțu-Romalo (1968: 143) identifies for the verb a tăia ‘to cut’.

(16) a. ta-i vs. b. tai-

ta-i tai-
ta-ie tai-e
tă-iem tăi-em
tă-iați tăi-ați
ta-ie tai-e

The two segmentations above mean to separate the stem from the ending, and
do not display a further segmentation in the ending sequences between affix and
desinence, something which Guțu-Romalo (1968) performs later on in her analysis.
Although she offers three advantages for the second segmentation, she ends up
picking segmentation a. over b., stating that verbs like a speria ‘to scare’ or a apropia
‘to bring near’ demand the first segmentation and, if verbs like a tăia get the same
segmentation, there will be a smaller number of ending patterns. This is shown in the
example below, together with the segmentation for the verb a umple ‘to fill’ and a ara
‘to plow’.

(17) tăia speria umple ara

t -i sp ri-i umpl-u ar-
t -i sp ri-i umpl-i ar-i
t -ie sp ri-ie umpl-e ar-ă
tă-i m speri-i m umpl-em ar-ăm
tă-i ți speri-i ți umpl-eți ar-ați
t -ie sp ri-ie umpl-u ar-ă

However, this type of argumention, as noted by Avram (1969b: 557), besides

being circular, is grounded on a false premise. There is no reason why a speria and a
apropia cannot receive the second type of segmentation or the first, (ibid.), and if they
do receive segmentation (16a) and thus have the same ending pattern as a tăia, it
won’t lead to a smaller number of ending sequences than it would lead if speria and
apropia were segmented, together with tăia, according to (16b), as shown in (18)
(with Guțu-Romalo’s fonetic transcriptions).

(18) tăia speria apropia

t i- sp rii- apr pii-
t i- sp rii- apr pii-
t i-e sp rii-e apr pi-e
tăi- m sperii- m apropii- m
tăi- ți sperii- ți apropii- ți
t i-e sp rii-e apr pii-e

Guțu-Romalo (1968: 145) further attempts to argue that the ending sequence
resulted from a common segmentation of verbs tăia, speria, and apropia according to
(16a) can be reduced to the sequence srufacing in verbs like ara, shown in (17).
Namely, –, –ă, and –ăm, she argues, can be shown to surface as –i, –e, and –em,
respectively, when preceded by a vocalic i as part of the stem. Apart from the
confusing fact that the i in tăia was already being argued by the same Guțu-Romalo
(1968: 143-144) to not be part of the stem, but of the ending, the same thing can be

said about the ending sequences resulted from the segmentation in (16b) (i.e. from the
point of view of phonological conditioning, it is irrelevant whether the triggering
context i is taken to be in the stem or in the ending as long as, in both cases, it
precedes the endings in question).
What her analysis does not attempt to account for, but more recent ones such
as Feldstein (2004)’s do, is the presence of –u as first person singular indicative
present ending in verbs such as umple in (17) and its link to other first person singular
ending surface forms. It has come to be a generally accepted fact that Romanian verbs
have a first person singular underlying marker which is deleted if preceded by only
one consonant: Chițoran (2002a: 43), Feldstein (2004: 182). This knowledge permits
us to unify the first person singular surface endings from (18) with the one in umple,
which represents a solid argument for segmentation (16b) and angainst (16a). This is
also a step forward in unifying all first person singular ending (surface) forms under
one underlying form (as long as the segmentation is assumed to be diffrent than what
Guțu-Romalo proposes), a thing which Guțu-Romalo’s analysis does not attempt to
do as systematically.
Another instance identified by Avram (1969b: 559) of ungrounded
delimitation between stem and ending is the inclusion of glides /y/ (in updated
analyses /j/) and /w/ in the conjugational ending, this leading, for instance, to
identifying three different (surface) forms for the imperfect inflectional marker (-a, -
ua, -ia) instead of just one (-a).
As mentioned previously, apart from the unfounded delimitation of stems from
endings with which Guțu-Romalo (1968) opens her structural analysis of the
Romanian verbal domain and which complicates the list of ending sequences with
either inaccurate or phonologically accountable surface forms, there are some other
problems with how the ending is further segmented and treated. She takes a
conjugational ending of the indicative present to be formed of an affix and a
desinence, Guțu-Romalo (1968: 149-150). The affix, linking the stem with the
desinence, is identified by the author based on the possibility of it being replaced by a
different affix to mark a different tense form. For instance, one can substitute –ă with
–a (as marker of the imperfect tense) in 1st person plural indicative present forms (and
not operate any changes in the stem or the desinence) to obtain the 1st person plural
forms of the imperfect (19a).

(19) a. cânt- -m ‘we sing’ vs. cânt- -m ‘we were singing’
vis- -m ‘we dream’ vs. vis- -m ‘we were dreaming’
b. cit- -m ‘we read’ vs. cit-e -m ‘we were working’
ven- -m ‘we come’ vs. ven-e -m ‘we were coming’
c. băt- -m ‘we beat’ vs. băt-e -m ‘we were beating’

Unlike the somewhat classical view represented by Graur (1957) in which

these affixes are seen as theme vowels, Guțu-Romalo (1968: 150) takes them to be
suffixes. However, the verbs in (19b) which have as theme vowel –i, do not seem to
undergo a simple substitution phenomenon where one present tense marker (suffix) is
switched with another tense marker. That is because the imperfect tense marker –a,
here, surfaces as the diphthong ea, which means a diphthongization took place
without an apparent trigger.
If, contrary to Guțu-Romalo’s stance, the –i in (19b) and all other affixes in
that position (between stem and person/ number desinence of the present tense) are
taken to be surface forms of the corresponding theme vowels, and if the status of -a in
the imperfect is maintained as a tense marker, then the diphthong -ea here is actually a
result of the interaction between these verbs’ theme vowel -i and the imperfect
marker. Chițoran (2002a: 206)’s representation of the Romanian diphthongs, quoted
in (20), supports this analysis.

(20) μ μ
Rt Rt

i a u a

a a
[ea] [oa]

In fact, both Chițoran (2002a) and Feldstein (2004) agree that the affixes
between stem and desinence (when they are phonologically realized) represent the
theme vowel, and Feldstein (2004: 183-184) also identifies the –a in the ending of the
imperfect as a tense marker, while he assumes the tense marker for the indicative
present to be null. So there is a diffrence in status between the indicative present affix

–i in (19b) and the imperfect tense affix –ea: the first is a realization of the theme
vowel, the second is a realization of a tense marker influenced by a theme vowel.
Another problem related to the analysis of these theme vowels/ affixes by
Guțu-Romalo (1968) has to do with the status of “extended suffixes” –ez and –esc and
their relation to thematic vowels –a and –i, respectively. Guțu-Romalo (1968: 151)
takes them all to have the same status of surface forms of the present tense suffix.
Avram (1969b: 558-562), wonders what the link is between –a, –ă, and –ez, on the
one hand, and –ez and –eaz, on the other, whether taken seperately or as sequences.
Namely, he asks whether they are phonologically or morphologically conditioned
allomorphs and expresses concern that neither of the two answers are valid: there is
no way a phonologically conditioned allomorph –ez could have two phonologically
conditioned allomorphs realizations, [ez] and [eaz], and a morphologically
conditioned allomorph {- , - , - z} could have morphologically conditioned
allomorph realizations. Some contexts in which these affixes appear are shown in (21)
and compared to the affix sequence {-ă, -a, -}.

(21) a lucra ‘to work’ vs. a ara ‘to plow’

lucr-ez ar-
lucr-ez-i ar-i
lucr-eaz-ă ar-ă
lucr-ă-m ar-ă-m
lucr-a-ți ar-a-ți
lucr-eaz-ă ar-ă

Chițoran (2002a:35) takes –ez and –esc to be “empty derivational suffixes”,

mere extensions of the stem which do however affect stress pattern. She takes them to
be part of the stem as theme vowels are, yet she does not think they can surface
together in the same verb form, Chițoran (2002a: 54). However, a uniform selection
between (verbs with) theme vowel –a and the affix –ez, on the one hand, and –i and
–esk, on the other, clearly exists. Feldstein (ND: 9) considers that these “extended
suffixes” may occur with or without theme vowel, and, when they do occur with, they
always carry stress and precede the (unstressed) theme vowel.
Finally, Avram (1969a:435-436) identifies problems in how Guțu-Romalo
defines homonymies and morphologically conditioned allomorphs, which are the
basis of her 10 conjugational classes, and expresses disaproval of her stem definition.

2.3 Feldstein’s segmentation of the Romanian verb

Departing in complexity from the segmentations presented so far by Guțu-

Romalo (1968) and even Chițoran (2002a), Feldstein (2004) posits a three-fold
segmentation of the conjugational ending of Romanian verbs. Table 16 below quotes
his analysis. The theme vowel, when it surfaces, is taken by Feldstein to be part of the
stem as does Chițoran (2002a).

Indicative present tense Imperfect tense

Tense Number Person Tense Number Person
1ag Ø Ø -u -a -m -u
2sg Ø Ø -i -a Ø -i
3sg Ø Ø Ø -a Ø Ø

1pl Ø -m -u -a -m -u
2pl Ø -t -i -a -t -i
3pl Ø -u Ø -a -u Ø

Table 16. Present and imperfect tense ending segmentation

An aspect in which Feldstein’s segmentation differs from Chițoran is the tense

marker. Chițoran (2002a: 43) assumes that –a in cânt-a-m ‘we were singing’ is the
theme vowel and not an imperfect marker and is thus part of the stem. Actually, she
doesn’t posit a tense marker for the present or imperfect, but she does for the
pluperfect which coincides with Feldstein (ND: 17)’s. His argument for –a being a
mark of the imperfect is relies on batem and băteam differing in exactly one element.

2.4 Conclusions so far

We have looked at the structural analysis of the Romanian verbal domain by

Guțu-Romalo (1968) and have seen the extent of its complexity and how the length of
surface form (allomorph) inventories as well as the number of underlying form
conjugational classes can be affected by improper segmentation of the verb and
improper analysis of the status of endings.
In what follows we will attempt to minimize the description by indentifying
more phonologically and morphologically conditioned allomorphs and examining and
offering extended and more rigorous accounts.


In what follows we will assume the more minimal segmentation of Chițoran

(2002a), and, when it will be required of us, we will further analyze the ending
sequence in the three-fold way introduced by Feldstein (2004), since the two
segmentations are not mutually exclusive, but rather the latter is a more detailed
updated version of the first. The phonemic vowel inventory is taken to be that in
Chițoran (2002b: 204) and given in (22), while the consonantal inventory shown in
(23) is from Chițoran (2002a: 10).

(22) Romanian vowel inventory:

i ɨ u
e ǝ o
ea a oa glides: j w

(23) Romanian consonant inventory:

labial dental palatal velar glottal
stops p, b t, d ʧ, ʤ k, g
fricatives f, v s, z ʃ, Ʒ h
nasals m n
approximants l

Chițoran (2002b: 204), along with others, treats the diphthongs as the low
counterparts of front and backed vowels. She argues for the phonemic status of some
occurrences of these two diphthongs by quoting minimal pairs (i.e. t mǝ ‘homework’
vs. te mǝ ‘fear’ and p ftǝ ‘appetite’ vs. po rtǝ ‘gate’), while other occurrences are
taken to be phonologically and morphologically conditioned. The glides, on the other
hand, are not considered phonemic, Chițoran (2002a: 95). Whether phonemic or
derived, the diphthongs are always under stress, Chițoran (2002b: 204). As mentioned
in section 2.2, Chițoran takes the diphthongs to share a syllable nucleus, and their
representation is quoted again in (24).

(24) μ μ
Rt Rt

i a u a

a a
[ea] [oa]

An interesting question arises as to the status of the first part of the

diphthongs, namely whether the initial elements are the same as the mid/back glides.
Chițoran (2002b)’s perception-production study revealed neutralization between back
glide-vowel sequence [wa] and the back diphthong [oa] (i.e. the underlying /ua/
sequences undergoing glide formation have the same phonetic realizations as [oa])
and a contrast between the mid glide-vowel sequence [ja] and the mid diphthong [ea]
which leads to [ja] receiving a different representation than that received by the
diphthongs, as shown below.

(25) σ

j a

The neutralization of [wa] and [oa] is explained by Chițoran (2002a: 248-249)

and Chițoran (2002b: 221) through a language specific difference in frequency, and
through the difficulty of maintaining a contrast between two back rounded glides
since the distance between the back vowels [o] and [u] is relatively smaller than the
one between front [i] and [e]. She also notes the asymmetry in the surface distribution
of glides [j] and [w], with the mid glide occurring in more contexts than the back one.
As stated several times before, Chițoran (2002a: 42-43) assumes the
conjugational suffixes of Guțu-Romalo (1968) to be realizations of theme vowels
(vowel of the traditional conjugational class) which are taken to be part of the stem
and not of the ending. This leads to the inflectional ending never carrying stress,
something which does happen for the imperfect tense if we assume the segmentation
of Feldstein, wherein – in the imperfect is taken to be a tense marker and not a theme
vowel, so it is part of the ending (which becomes stressed) and not of the stem.

However, since our interest is only in the indicative present, and in the indicative
present Feldstein posits a null tense marker with the theme vowel surfacing before the
ending (so as final part of the stem), the two segmentations are more or less
equivalent in relation to the present study.

3.1. Hiatus resolution

Romanian tolerates surface hiatus in a small number of lexical items, bearing

mid-mid (i.e. /e e/, /o e/, /e o/, and /oo/), mid-low (i.e. /e a/, /o a/) or low-mid (i.e. /a
e/, /a o/) sequences, Chițoran (2002a: 121-122). For the majority of underlying hiatus
contexts, though, the situation is resolved either through glide epenthesis or glide
formation, as shown in Chițoran (2002a: 96-97) and quoted below, depending on
stress placement.

(26) glide epenthesis glide formation

ha.j .nǝ ‘mean’ F h ij.nǝ ‘coat’
v .je ‘vineyard’ vij r.me ‘worm’
l .w .dǝ ‘praise’ no item
pǝ.w n ‘peacock’ no item

Example (26) above shows that glide formation does not occur in the case of
the back high vowel /u/, regardless of stress location. Chițoran (2002a: 103)’s analysis
shows that epenthetic [j] surfaces when the second vowel in the sequence is front, or
when the first vowel of the sequence is /i/ and the second one is not /u/, and epenthetic
[w] surfaces either when the second vowel in the sequence is /u/, or when the second
element is back (but not /u/) and the first is not /i/.

3.2 Regressive metaphony, diphthongization, and deletion

According to Chițoran (2002b: 204), the two diphthongs in Romanian and the
vowel /a/ are subject to (regressive) metaphony, the process by which the vowel of an
inflectional marker affects the height of the stressed vowel of the stem. Actually,
Chițoran (2002a: 98) takes the majority of vowel alternations in Romanian to involve
height. She identifies three types of alternations that occur under stress in Romanian:
the two diphthingizations, e ea, o oa, and ǝ ~ a. Chițoran (2002a: 206) argues that
Romanian diphthongization should be treated as a vowel lowering under stress. The
third alternation of stressed [a] with unstressed [ǝ] should not be considered vowel

reduction, like in English, since shwa is not a phonologically reduced vowel and has
phonemic status in Romanian. In the context of stress, the diphthongs always surface
in 3rd person singular (and sometimes plural) forms (27a), while the other forms in the
paradigm contain the mid vowels. Newer verbs, however do not display the
diphthongs (27b).

(27) a. dorm ‘I sleep’ / do rme ‘s/he/they sleep’

skol ‘I awake’ / s o lǝ ‘s/he/they awake’
lucr- z ‘I work’ / l cr-e z-ǝ ‘s/he/they work’
b. implor ‘I implore’ / implor- ǝ ‘s/he/they implore’

As far as vowel alternations taking place outside of stress, Feldstein

(2004: 185-186) identifies three processes:
1. any unstressed front-vowel theme (regardless of frontness or backness)
becomes mid before an unstressed desinential vowel;
2. an unstressed theme vowel becomes mid before a word-final vowel or zero
3. an unstressed mid vowel is deleted when it precedes an unstressed
desinential vowel; an unstressed high desinential vowel is deleted when it
is preceded by a non-mid (high or low) vowel.
These do not apply to verbs with monosyllabic stems, where the theme vowel is
stressed and the first person marker –u, instead of being deleted, surfaces as a glide.

3.3 Palatalization

Chițoran (2002a: 173), along with others, takes palatalization to be triggered

by an underlying /i/, Spinu (2007: 306). Word final underlying high vowels surface as
glides if they are preceded by a vowel. If they are preceded by a consonant, /u/ is
deleted while /i/ surfaces as palatalization of the preceding consonant, affecting
secondary articulation. Final –i can also affect primary place of articulation for a
preceding coronal or velar consonant, in which case the palatalization is
morphologically conditioned and applied only at morpheme boundary. The author
finds consonantal alternations triggered by palatalization (those already listed in
section 2.1.4) to be regular, with the exception of palatalized /l/ which can either
surface as [j] or as [lj], but occurs only in the nominal domain (which is why it’s not
listed in table 12).

In order to account for these alternations, Chițoran (2002a: 186) first adds the
feature [strident] for coronal segments to capture the difference between /t/ and /ʦ/,
treating affricates as [+strident] stops. The triggers for morphologically conditioned,
feature changing palatalization are –i for coronals and velars and –e for velars.
Chițoran (2002a: 189) assumes palatalization to generally be the spreading of the
coronal feature from a front vowel to the V-place of a preceding consonant which
leads to secondary articulation, and redundantly specifies front vowels as [–anterior].
In the case of coronal and velar obstruents (/s/, /z/, /k/, /g/), palatalization also
involves a stage, for Chițoran, in which the coronal or dorsal articulation is affected
and secondary articulation features are promoted to primary articulation. Namely,
[–anterior] of –i is spread to C-place [coronal] for coronals /s/ and /z/ which turn into
[ʃ j] and [Ʒj] under palatalization, respectively, and [coronal] spreads to C-place and
[dorsal] is delinked for velars /k/ and /g/ which result in [ʧ j] and [ʤj]. For coronal
fricatives only the [–anterior] is promoted to primary articulation. For coronal stops,
palatalization involves the insertion of [+strident] along with vowel feature spreading.
Chițoran’s accounts for alternations st-ʃt j and sk-ʃt j are given below.

(28) st-ʃt j
C-Place C-Place

[coronal] [coronal] vocalic



(29) sk-ʃt j
C-Place C-Place

[coronal] vocalic




According to Chițoran, –u for verbs is the first person singular marker in the
indicative present tense and –i, also for verbs, is the second person singular marker.
However, according to Feldstein (2004: 183), who posits separate markers for number
and person, –u and –i are 1st and 2nd person markers independent of number and tense.
This correctly predicts that they should be contained by 1st and 2nd person plural
endings as well. Surely enough, the 2nd person plural present tense ending desinence
-ʦ j appears to be the result of palatalization of –t, which Feldstein (2004) takes to be
the 2nd person plural number marker (i.e. preceding the person marker). This means
that Feldstein’s segmentation, at least as far as indicative present number and person
goes, is on the right track.

3.4 Feldstein’s derivation of indicative present forms

Feldstein (2004: 185) assumes that the conjugated form of a verb is derived in
cycles corresponding to the affixation of the marker. The process is finished once the
ending is fully attached to the stem and all phonological rules have been applied.
Thus, at each cycle of conjugation (stem + tense, stem + number, and stem + person),
the rules stated in section 3.2 will be applied in order to change theme vowels to mid-
vowel height and then to delete vowels in the sequence of theme vowel plus vocalic
desinence. Table 17 shows how the first person singular of the indicative present is
derived for three verbs.

Cycle a invita a bate a sări Comments

1. Basic input invita+Ø-Ø-u# bate+Ø-Ø-u# sari+Ø-Ø-u#
2. Tense invita+Ø-u# bate+Ø-u# sari+Ø-u# Elimination of zero
tense morpheme
3. Number invita+u# bate+u# sari+u# Elimination of zero
number morpheme
4. Person (a) invitǝ+u# -- sare+u# Unstressed theme
vowel → mid
preceding word-final
(b) invit+u # bat+u# sar+u# Deletion of unstressed
mid [e/ǝ] before
desinential vowel
5. Postlexical invit bat sar Loss of -u# unless
blocked due to
preceding [consonant
+ liquid] sequence

Table 17. Derivation of 1st person singular present forms


In what follows we will look again at Guțu-Romalo (1968)’s ending

sequences, this time to see if the list can be reduced based on the information we’ve
drawn in the previous chapter about phonological processes occurring in Romanian.
We will assume the affixes Guțu-Romalo (1968) considers as suffixes before
desinences to be surface forms of the theme vowel. The theme vowels are assumed to
be –a, –e, –i and –î. Verbs in –e can and will be treated as having theme vowel –e,
with the stress falling on the theme vowel instead of the root. Feldstein (ND: 8) argues
for this interpretation due to the already discussed diphthongization that e undergoes
under stress. The ending –î can be treated as a subclass of –i, however, for ease of
argumentation, we will treat them separately.
Since alternations in the consonantal or vocalic, or both, segments of the stem
can be accounted through the phonological processes discussed in the previous
chapter in a more trivial matter than the ending sequences, and since the ending
sequences are more numerous than the stem alternations, we will only look at ending
sequences as see if they can be reduced due to phonological or morphological

4.1 Reducing the ending sequences

Looking at Guțu-Romalo’s (1968: 196-197) 38 ending sequences for the

indicative present (quoted in appendix A), we first see that 15 of these refer to verbs
with –a as theme vowel, 7 with –i, 2 with –î, and 11 with –e. Sequences 36 and 37 are
listed with a null theme vowel/ infinitive ending, but they represent aberrant/ irregular
verbs ști ‘know’ and fi ‘be’, so they will be ignored. Sequence 22 is listed as having
–a as theme vowel/ infinitive ending, yet it is shown to represent verbs plăcea ‘like’
and tăcea ‘be silent’, both of which actually end in –e and display the theme vowel
–e in the ending pattern as well so they will be treated as – theme vowel verbs. Since
patterns 31 to 38 are considered by the author herself to represent irregular or aberrant
verbs, these will be treated last.

4.1.1 Verbs with theme vowel (stressed or unstressed) -e

As discussed in section 2.1.3, patterns 21 to 30, meaning all patterns
representing verbs which have –e as theme vowel, appear to be surface forms of an
underlying pattern. The 4 surface forms are shown in (28).
(28) a. Ø+Ø b. Ø+Ø
Ø + [-i] Ø + [-i]
Ø +[-e] Ø +[-e]
[- -]+[-m] [-e-]+[-m]
[- -]+[-ț ] [-e-]+[-ți]

c. Ø +[-u] d. Ø +[-u]
Ø + [-i] Ø + [-i]
Ø +[-e] Ø +[-e]
[-e-]+[-m] [-ie-]+[-m]
[-e-]+[-ți] [-ie-]+[-ți]
Ø +[-u] Ø +[-u]

Form (28a) is represented by patterns 21 and 22 which both describe verbs

ending in –e , (i.e. have – as theme vowel), while (28b) is represented by patterns
23, 25, 26, 27, 29, and 30, which describe verbs with the infinitive ending in –e (i.e.
have –e as theme vowel). The only diffrence in these first two surface forms is the
stressed/ unstressed [-e-] in the 1st and 2nd person plural which can be explained by the
diffrence between theme vowel stress: (28a) represents verbs with stressed –e, while
(28b) represents verbs with unstressed –e.
Forms (28c) and (28d) represent pattern 24 and 28 respectively, and appear to
differ among each other only in 1st and 2nd person plural forms: (28d) has [-ie-] as
“suffix”, while (28c) has only [-e-] on the same position. Since the theme vowel is –e
for both of these patterns, and since form (28d) (i.e. ending pattern 28) represents
verbs whose infinitive ends in –ie (i.e. scrie ‘write’) while the other represents verbs
whose infinitives end just in –e, I argue that the two forms are identical due to i in
[-ie-] being part of the invariable stem and not a specific realization of the theme
vowel (which we also asume to be part of the stem but which is generally subject to
alternation). This means that Guțu-Romalo (1968: 196-197) wrongly posited pattern
28 to be diffrent than 24, due to a wrong separation of the stem from the endings (i.e.
the segmentation is scri-e instead of scr-ie). There is thus no diffrence between forms
(28c) and (28d), since verbs represented by pattern 28 are actually also represented by
pattern 24.

What is left now is to show how pattern (28b) and (28c) are connected. These
two patterns differ in 1st person singular and plural and 2nd person singular. We have
already mentioned in this chapter that Chițoran (2002a) and Feldstein (2004: 183)
take an underlying /u/ to be a 1st person marker, which only surfaces when following a
consonat cluster. Form (28c) (and d) displays this in 1st person singular and plural and
it does so because it represents verbs which end in a consonant cluster (more
precisely, stop+liquid cluster), while form (28b) (and a) does not display this
underlying /u/ because it does not represent verbs whose stems end in a stop+liquid
cluster. As for the diffrence in 2nd person singular forms of the two ending sequences,
it is clear that the already assumed underlying 2nd person marker /i/ surfaces as a
vocalic i in one context and palatalization in the other context.
If we then assume Ø+[-e] in the 3rd person singular of all three possible surface
forms of the 10 ending sequences identified by Guțu-Romalo (1968) to actually be
[-e-]+Ø instead (i.e. the –e is actually the theme vowel and not a 3rd person singular
marker), we can say that the 3 surface forms fit into the segmentation and ending list
proposed by Feldstein (2004: 183), with the “suffixes” being the theme vowel
undergoing the processes defined in section 3.2 and exemplified in 3.4. This is shown
in (29), with the hyphen delimiting in this order: stem, theme vowel, tense marker,
number marker, person marker.

(29) stem-e-Ø-Ø-u

So far, we’ve offered an account that reduces Guțu-Romalo (1968: 196-197)’s

10 ending sequences for verbs with infinitives in –e to one underlying form, and 3
surface forms.

4.1.2 Verbs with theme vowel –i and –î

We now turn our attention to the 7 types of verbs having theme vowel –i. We
delimit those patterns that represent verbs which receive the stem extension –esc–
from those representing verbs which receive only the theme vowel –i. Thus, we have

patterns 13-16 employing the theme vowel –i and 17-19 employing the –esc–
Patterns 17 to 19 display only 2 surface forms given in (30), where 17 and 19
share (30a) and 18 is represented by (30b).

(30) a. [- sc-]+Ø b. [-i sc-]+Ø

[- șt-]+[-i] [-i șt-]+[-i]
[- șt-]+[-e] [-i șt-]+[-e]
[- -]+[-m] [- -]+[-m]
[- -]+[-ți] [- -]+[-ți]
[- sc-]+Ø [-i sc-]+Ø

The only diffrence between forms (30a) and (30b) lies in an additional i (actually an
[j]) that precedes the extended suffix –esc–, when this suffix surfaces. This diffrence
can be explained if we look at what verbs are given by Guțu-Romalo (1968: 196-197)
to represent pattern 18, namely alcăt i ‘compose’. Although the verbs we are
discussing now have –i as theme vowel, since we assumed the theme vowel to surface
after the extended suffix, the i before –esc– in pattern 18 (or form (30b)) is not a
realization of the theme vowel, but rather an epenthetic glide resulted from the
resolution of the hiatus between the final u of the stem (based on segmenting the
infinitive as alcăt -i) and the e of the extended suffix. Thus, patterns 17-19 are argued
to share the same underlying pattern given in (31) in the same style as in (29).

(31) stem-esc-i-Ø-Ø-u

Regarding the positive vs. negative realization of the extended suffix,

Feldstein (ND:9) notes that when it is unstressed it gets deleted and only the theme
vowel surfaces. This is what happens for 1st and 2nd person plural forms of verbs
which receive any of these stem extensions (i.e. –ez, –esc, –ăsc).
Regarding the status of [- sk-] and [- șt-], in the light of the alternations
discussed in chapter 3 (particularly palatalization of cluster sk), it is clear that the
second is a morphologically conditioned (by the following 2nd person marker –i) but
phonologically restricted allomorph of the first. Similarly, in the light of the
diphthongization process discussed in 3.2, [-e z-] is the morphologically conditioned
but phonologically restricted allomorph of [-ez-].
A further argument to confirm that what appears as 3rd person singular
desinence in Guțu-Romalo (1968)’s analysis is actually a realization of the theme
vowel, which indirectly confirms Feldstein (2004)’s segmentation and identification
of tense, number, and person markers (in the indicative present, at least), is that the
presence of [-e] in the 3rd person singular form of verbs having the theme vowel –i
(with or without –esc–) can be described as a lowering of the high theme vowel to
mid –e. Feldstein (2004: 184) argues that the theme vowel either gets lowered or
raised to reach mid e or ǝ in the 3rd person. If it were not tied to the theme vowel (i.e a
realization of it) and the final surface segment in the 3rd person desinence were a
person marker, then there would be no way to analyze it, no way to determine which
is the underlying form.
Looking now at the patterns representing the verbs with the theme vowel –i
but without –esc– (patterns 13-16), we see that patterns 15 and 16 coincide (32a),
while 13 and 14 differ in a few forms. (32b) represents pattern 14 and (32c) represent
pattern 13.

(32) a. Ø+ Ø b. Ø+ [-i] c. Ø+ Ø
Ø+ [-i] Ø+ [-i] Ø+ [-i]
Ø+ [-e] Ø+ [-ie] Ø+ [-ă]
[- -]+[-m] [- -]+[-m] [- -]+[-m]
[- -]+[-ți] [- -]+[-ți] [- -]+[-ți]
Ø+ Ø Ø+ [-ie] Ø+ [-ă]

The 3rd person singular and plural forms of (32b) can be accounted as the
result of hiatus resolution between the back vowel u (contribu-i) of the stem and the
high front theme vowel which was lowered to e giving rise to an epenthetic glide.
More interesting is the presence of –ă in 3rd person singular/ plural in (32c). It may be
accounted as a result of dissimilation between the vowel of the preceding syllable (e
in acoper-i ‘cover’) and the lowering into –e of the theme vowel –i. So the lowering
does occur but results in the mid center ǝ instead of the mid front e.

The two patterns for verbs with theme vowel –î, 12 and 20, represent verbs
which either take or don’t take –ăsc– as an extension of the stem. The distinction
between verbs which take –î instead of –i was initially phonologically conditioned,
Feldstein (2004: 11) and Costanzo (2011: 97), but is now lexical. I also take the
distinction of verbs which do and don’t get extended suffixes to be lexically
conditioned as well.

4.1.3 Verbs with theme vowel –a

Turning now our attention to the final group of verbs, those which have –a as
theme vowel, we further split these into those which accept extended suffix –ez–
(patterns 7-11) and those which do not (patterns 1-6). For the first group (with –ez–),
we notice pattern 11 is no longer in accordance with the contemporary standard
language (i.e. verbs like crea ‘create’ maintain the hiatus and do not resolve it with an
epenthetic glide), so we will eliminate this pattern and treat its verbs as conjugating
with patterns 7 or 8 (they are identical). The surface forms are given in (33):

(33) a. [- z-]+ Ø b. [-i z-]+Ø c. [- z-]+Ø

[- z-]+ [-i] [-i z-]+ [-i] [- z-]+[-i]
[-e z-]+[-ă] [-i z-]+[-ă] [- z-]+[-ă]
[- -]+[-m] [-i -]+[-m] [- -]+[-m]
[- -]+[-ți] [-i -]+[-ți] [- -]+[-ți]
[-e z-]+[-ă] [-i z-]+[-ă] [- z-]+[-ă]

In the case of (33b), it is clear that the –i preceding all suffixes (whether the
extended or the theme vowel) is not an epethentic glide but part of the invariable
stem. This leads to forms (33b) and (33c) to appear identical. Form (33c),
representing ending pattern 10 for which veghea ‘look after’ is given as an example,
only appears to not display the diphthongization of – z into –e z in the 3rd person,
which is displayed by (33a) and avoided by (33c), I argue, due to the preceding
semivocalic i in the stem which blocks the diphthongization. In reality, contrary to the
posited – z by Guțu-Romalo (1968), the verb does display –e z in the 3rd person as
can be seen in (34) and as is also identified by Barbu (2009: 112).

(34) vegh- z
vegh- z-i
vegh-e z-ă
vegh- -m
vegh-e -ți
vegh-e z -ă

The final diffrence between (33a), on the one hand, and (33b) and (33c), on the
other, is the [-ǝ-] vs. [- -] in the 1st person plural. In (33a) and (33b), a raising of the
theme vowel can be argued to have occurred, where the theme vowel in (33b) is
morphologically conditioned (but phonologically restricted) to surface as mid front [e]
instead of mid mid [ǝ], since it is closer to the high front i of the stem. In (33c),
however, the e can be argued to be from the stem and not the raising of the theme
vowel –a to front mid. Once more, the –ă in the 3rd person singular is a clear raising
of the theme vowel under loss of stress as discussed in section 3.2.
Regarding the verbs in –a which do not receive extended – z, patterns 3 and 6
(see appendix A) display a preceding i and u which come from the invariable stem
rather that from some unknown interaction with the surfacing theme vowel. We also
see a difference in surface form of the low theme vowel in 3rd person singular and 1st
person plural when it is raised to mid. The surfacing of theme vowel –a as mid front
[e] in pattern 6 can be argued, as we’ve done previously, to be morphologically
conditioned and phonologically restricted by the underlying preceding front high
vowel i from the stem, which surfaces as a glide to resolve hiatus.

4.2 Proposing a final division

Based on what we have seen and discussed so far, it becomes evident that the
3-fold ending segmentation together with the specific markers proposed by Feldstein
(2004) (for the indicative present, at least) is the best (underlying) representation of
Romanian ending sequences proposed so far. The only remaining descriptive division
that should be made in the verbal domain is the one given below. It is in tune with the
extended traditional one, only that this time the alternations have been accounted for.

(35) stem- stem- st m-e stem- stem-

stem- z-a stem- sc-i stem- sc-î


In this work, I have looked at the Romanian verbal paradigmatic domain and
shown the extent of its complexity, by quoting the many results in the ample structural
analysis of Romanian morphology conducted by Guțu-Romalo (1968).
Employing contemporary insights into the phonology of Romanian given
mainly by Chițoran (2002)’s works and arguing for Feldstein (2004)’s segmentation
of the Romanian verb, we’ve seen that the majority of vocalic and consonantal
alternations in the Romanian verbal domain, identified by previous linguists such as
Lombard (1955), Moisil (1960), and Guțu-Romalo (1968) and listed by us in the
second chapter and the appendinx, are in fact regular and phonologically or
morphologically conditioned, so an account can be given and rules of formation can
be postulated. These postulated rules then render unnecessary stem representations
such as the ones proposed by Moisil (1960) or Guțu-Romalo (1968) which make use
of additional symbols to mark variable letters.
These observations have enabled me to give an account for the many diffrent
ending sequences which occur in the indicative present and attempt a reduction of the
38 sequences identified by Guțu-Romalo (1968) which was not based on the various
homonymies employed by her but on underlying phonological representations, an
approach in tune with Feldstein (2004)’s proposed cyclic derivation.
After this reduction, the analysis presented in this work came full circle in
recognising that the division which best fits the Romanian verb paradigm is one in
tune with the extended traditional, where the sole classification criterion is the lexical
one which is able to account for the selection of one theme vowel (and non-thematic
stem extension) over another.
Thus, we have shown that, although at a second glance Romanian morphology
appears to be complex and rich with alternations, these alternations are mostly regular
and phonologically accountable.


Avram, A. 1969a. Pe marginea unei morfologii structurale a limbii române I. Studii și

cercetări lingvistice XX (4): 433-453.
Avram, A. 1969b. Pe marginea unei morfologii structurale a limbii române II. Studii
și cercetări lingvistice XX (5): 557-577.
Avram, M. 2001. Gramatica pentru toți. Bucharest: Humanitas Educațional.
Barbu, A.-M. 2009. Conj garea verbelor românești. Bucharest: Editura Coresi.
Chițoran, I. 2002a. The Phonology of Romanian: a constraint-based approach, New
York: Mouton de Gruyter.
Chițoran, I. 2002b. A perception-production study of Romanian diphthongs and glide-
vowel sequences. Journal of the international phonetic association, 32/2, 204-222,
United Kingdom: International Phonetic Association.
Constantinescu-Dobridor, G. 2001. Gramatica limbii române. Bucharest: Editura
Didactică și Pedagogică
Costanzo, A. R. 2011. Romance conjugational classes: learning from the peripheries.
PhD thesis, Ohio State University.
Dinu, L. P., Niculae, V., Șulea, O.-M. 2012. Learning how to conjugate the Romanian
verb. Rules for regular and partially irregular verbs. In W. Daelemans, M. Lapata, and
L. Màrquez (eds.), EACL 2012, 13th conference of the european chapter of the
association for computational linguistics, 524-529. Stroudsburg: The Association for
Computational Linguistics.
Feldstein, R. F. ND. Romanian Verb Handbook. Indiana University.
Feldstein, R. F. 2004. On the structure of syncretism in Romanian conjugation. In J.
Auger, J. C. Clements and B. Vance (eds.), Contemporary Aproaches to Romance
Linguistics. Selected Papers from the 33rd linguistic symposium on Romance
Languages, 177-195. Amsterdam/Philadelphia: John Benjamins.
Felix, J. 1964. Classification des verbes roumains. Philosophica Pragensia, VII, 291-
Guțu-Romalo, V. 1968. Morfologie str ct rală a limbii române. Bucharest: Editura
Academiei Republicii Socialiste România.
Graur, A. 1957. Note asupra structurii morfologice a cuvintelor. Studii de gramatică,
II, 3-18, Bucharest: Editura Academiei.
Juilland, A., Edwards, P. M. H., and Juilland, I. 1965. Frequency dictionary of
Romanian words. The Hague: Mouton.
Juilland, A. and Edwards, P. M. H. 1971. The Romanian verb system. The Hague:
Lombard, A. 1955. Le verbe roumain. tude morphologique, volume 1. Lund:
Lombard, A. and Gâdei, C. 1981. Dictionnaire morphologique de la langue roumain.
Bucharest: Editura Academiei Republicii Socialiste România.
Moisil, G. 1960. Probleme puse de traducerea automată. Conjugarea vebelor în limba
română. St dii și cercetări lingvistice XI (1): 7-25.
Papastergiou, I., Papastergiou, N., and Mandeki, L. 2007. Verbul românesc - reguli
pentru înlesnirea însușirii indicativului prezent. Romanian national symposium
”Directions in romanian philological research”, 7th edition.
Spinu, L. 2007. Perceptual properties of palatalization in Romanian. In J. Camacho,
L. Sanchez, N. Flores-Ferran, V. Deprez, and M. Jose Cabrera (eds.), Romance

Linguistics 2006: Selected papers of the 36th linguistic symposium on Romance
Languages (LSRL), 303-319. Amsterdam: John Benjamins.
Sanchez Miret, F. 2011. Câteva observaţii asupra diftongării metafonice în limba
română. In O. Felecan and D. Felecan (eds.), Confl enţe lingvistice şi filologice.
Omagi profesor l i Nicolae Felecan la împlinirea a 70 de ani, 97-109.
Cluj-Napoca: Mega.
Tiktin, H. 1905. R mänisches Elementarbuch. Heidelberg: C. Winter.


A. Guțu-Romalo (1968)’s 38 ending sequences for the indicative present

Tense 1 2 3 4 5
Infinitive [- ] [- ] [-u ] [-u ] [- ]
Present Ø+Ø Ø+[-u] Ø+ [-u] Ø+ Ø
indicative Ø+[-i] Ø+[-i] Ø+ [-i] Ø+ [-i]
Ø+[-ă] Ø+[-ă] Ø+ [-uă] Ø+ [-ă]
[- -]+[-m] [- -]+[-m] [-u ]+[-m] Ø+[-u ] [- ]+[-m]
[- -]+[-ți] [- ]+[-ți] [-u ]+[-ți] [- ]+[-ți]
Ø+[-ă] Ø+ [-ă] Ø+[-uă] Ø+ [-ă]
Example ara afla continua plouă preceda

6 7 8 9 10
[-i ] [- ] [-u ] [-i ] [- ]
Ø+ [-i] [- z-]+ Ø [- z-]+Ø [-i z-]+Ø [- z-]+Ø
Ø+ [-i] [- z-]+ [-i] [- z-]+ [-i] [-i z-]+[-i] [- z-]+[-i]
Ø+ [- e] [-e z-]+ [ă] [-e z-]+[-ă] [-i z-]+[ă] [- z-]+[-ă]
[-i ]+[-m] [- -]+[m] [- -]+[-m] [-i -]+[-m] [- -]+[-m]
[-i ]+[-ți] [- -]+[ți] [- -]+[-ți] [-i -]+[-ți] [- -]+[-ți]
Ø+ [-ie] [-e z-]+[ă] [-e z-]+[-ă] [-i z-]+[-ă] [- z-]+[-ă]
apropia lucra perpetua sublinia veghea

11 12 13 14 15 16
[-(i) ] [-î] [- ] [- ] [- ] [- ]
[- z-]+Ø Ø+ Ø Ø+ Ø Ø+ [-i] Ø+ Ø Ø+ Ø
[- z-]+ [-i] Ø+ [-i] Ø+ [-i] Ø+ [-i] Ø+ [-i] Ø+ [-i]
[-e z-]+[-ă] Ø+[-ă] Ø+ [-ă] Ø+ [-ie] Ø+ [-e] Ø+ [-e]
[-i -]+[-m] [- -]+[-m] [- -]+[-m] [- ]+[-m] [- -]+[-m] [- -]+[-m]
[-i ]+[-ți] [- -]+[-ți] [- -]+[-ți] [- ]+[-ți] [- -]+[-ți] [- -]+[-ți]
[-e z]+[-ă] Ø+ [-ă] Ø+ [-ă] Ø+[-ie] Ø+ Ø Ø+ Ø
crea coborî acoperi contribui sări fugi

17 18 19 20 21 22
[-i] [-i] [-i] [-î] [-e ] [- ]
[- sc-]+Ø [-i sc-]+Ø [- sc-]+Ø [- sc-]+Ø Ø+Ø Ø+Ø
[- șt-]+[-i] [-i șt-]+[-i] [- șt-]+[-i] [- șt-]+[-i] Ø +[-i] Ø + [-i]
[- șt-]+[-e] [-i șt-]+[-e] [- șt-]+[-e] [- șt-]+[-e] Ø +[-e] Ø +[-e]
[- -]+[-m] [- -]+[m] [- -]+[-m] [- -]+[-m] [- -]+[-m] [- -]+[-m]
[- -]+[-ți] [- -]+[ți] [- -]+[-ți] [- -]+[-ți] [- -]+[-ți] [- -]+[-ți]
[- sc-]+Ø [-i sc-]+Ø [- sc-]+Ø [- sc-]+Ø Ø+Ø Ø+Ø
porni alcăt i lovi hotărî putea tăcea

23 24 25 26 27 28
[- ] [- ] [- ] [- ] [- ] [-i ]
Ø+Ø Ø +[-u] Ø+Ø Ø+Ø Ø+Ø Ø+[u]
Ø + [-i] Ø + [-i] Ø + [-i] Ø + [-i] Ø + [-i] Ø+[-i]
Ø +[-e] Ø +[-e] Ø +[-e] Ø +[-e] Ø +[-e] Ø+[-ie]
[-e-]+[-m] [-e-]+[-m] [-e-]+[-m] [-e-]+[-m] [-e-]+[-m] [-ie-]+[-m]
[-e-]+[-ți] [-e-]+[-ți] [-e-]+[-ți] [-e-]+[-ți] [-e-]+[-ți] [-ie-]+[-ți]
Ø+Ø Ø +[-u] Ø+Ø Ø+Ø Ø+Ø Ø+[-u]
începe umple face aprinde alege scrie

29 30 31 32 33 34
[- ] [- ] [-u ] [- ] [-e ] [-e ]
Ø+Ø Ø+Ø Ø+[-u] Ø+[-u] Ø + [-u] Ø +[-u]
Ø +[-i] Ø +[-i] Ø+[-i] Ø+[-i] Ø + [-i] Ø + [-i]
Ø +[-e] Ø +[-e] Ø+Ø Ø+[-ă] Ø+ Ø Ø+ Ø
[-e-]+[-m] [-e-]+[-m] [-u -]+[-m] [- -]+[-m] [- -]+[-m] [- -]+[-m]
[-e-]+[-ți] [-e-]+[-ți] [-u -]+[-ți] [- -]+[-ți] [- -]+[-ți] [- -]+[-ți]
Ø+Ø Ø+Ø Ø+[-u] Ø+[-u] Ø + [u] Ø+Ø
rupe coace lua da bea vrea

35 36 37 38
[-e ] Ø Ø [- ]
Ø +[-m] Ø +[-u] Ø+Ø Ø+ [-u]
Ø + [-i] Ø + [-i] Ø + [-i] Ø+ [-i]
Ø+ [-e] Ø+ [-ie] Ø+ [-e] Ø+ [-ă]
[- -]+[-m] Ø +[-m] [- -]+[-m] [- ]+[-m]
[- -]+[-ți] Ø +[-ți] [- -]+[-ți] [- ]+[-ți]
Ø + [-u] Ø + [-u] Ø+Ø Ø+[-u]
avea ști fi preda

B. Interactions between consonantal and vocalic alternations for the indicative
present, as identified by Guțu-Romalo (1968)

Vocalic  ă/ / î/ /e /o ă/ ă/e /ă /e / /o
Consonantal  /u
[k] / [ʧ] + + + +
[g] / [ʤ] + + +
[t] / [ʦ] + + + + + +
[d] / [z] + + + + +
[s] / [ʃ] + + + +
[s (t)] / [ʃ (t)]
[(ʃ) k] / [(ʃ) t] + + +
[(s) k] / [(ʃ) t]
b/ +

C. Guțu-Romalo’s classification according to number of stem allomorphs and
syncretism. Indicative present reduced and recompiled version

Stem New Old Verb

allomorphs subclass subclasses Stem syncretism Examples total

I 1 I,II,VI; s1=s3=s4=s5=s6, s2 acorda ‘award’: acord/z 79

3 XII; 4 III
II 1 III,VIII. s1=s2=s4=s5, s3=s6 așez ‘I lay down’, așeaz 10
III 1 IV. s1=s2=s3=s6, s4=s5 crăpa ‘crack’: crap, crăp 13
IV 1 VII. s1=s2=s4=s5=s6, s3 dorm ‘I sleep’, doarm 3
V 2 XVI, s2=s3=s4=s5, s1=s6 fugi ‘run’: [fug], [fuʤ] 38
3 IX,X;
4 III; 5 II.
VI 1 X; 2 XX s1=s3=s6, s2=s4=s5 auzi ‘hear’: aud, auz 7
VII 1 XI s2=s3=s4=s5=s6, s1 sufăr ‘I suffer’, sufer- 1
I 2 I, II, IV; s1=s4=s5, s3=s6, s2 cert ‘I scold’, cerț-/ceart- 19
3 I.
II 2 V, VI; s1=s3=s6, s4=s5, s2 las ‘I leave’, lăs-/laș- 26
3 3 II, VI
III 2 VII s1=s2, s3=s6, s4=s5 scol ‘I awake’, scoal/scul 9
IV 2 VIII s1=s2=s6, s4=s5, s3 mor ‘I die’, moar-/mur- 2
V 2 IX, s1=s4=s5=s6, s2, s3 șed ‘I sit’, șez-/șad- 3
VI 2 XIX. s1=s6, s2=s4=s5, s3 slobod ‘I free’, sloboz-/ 1
VII 3 VII,VIII, s1=s6, s3=s4=s5, s2 cos ‘I sew’, coș-/ coas- 9
XI; 4 II,
IV; 5 I
I 3 III; 4 I. s3=s6, s4=s5, s1, s2 purta 2
II 3 IV s1=s6, s4=s5, s2, s3 putea 1